The ultimate guide to motherboards

If you need to future proof your system it might be worth upgrading to DDR3 but be prepared for a heart-stopping price tag

By God it's a scary time to be buying a motherboard. We've gotten so comfortable with Intel's age-old LGA775 socket – which has been with us since 2005's Pentium 4, amazingly – that the bad old days of deciphering which chip would go in which board seem like a distant nightmare.

Well, prepare to start screaming and waking up in a cold sweat again, because right now there's a confluence of confusing sea-changes going on. These mean picking up a new mobo involves either sacrificing future-proofing or having to drop a small fortune on a new chip and RAM as well.

Yes, it's horribly unfair, but please cease freaking out and allow us to guide you through this new motherboard maze: it's not as convoluted as it looks. Chief perpetrators of this technotragedy are new-ish memory type DDR3 and Intel's newer still Core i7 CPU.

DDR2 has done us proud for a few years, but it's finally on its way out, DDR3 having lately become the new standard for both Intel's Core 2 and AMD's Phenom II chip ranges. While the day-to-day performance boost from pairing DDR3 with a Core 2 or Phenom 2 is pretty much negligible, picking up DDR2 sticks and/or a DDR2 mobo today shuts the door on future upgrade options.

On the other hand, grabbing a DDR3 board will most likely involve having to grab some DDR3 too. It's fortunate then that the memory type's once-scandalous cost has recently fallen off a cliff, but picking up 4GB will still set you back at least £60.

Core i7

The Core i7, meanwhile, is a real barnstormer of a chip, offering around a 25 per cent performance boost over Core 2 chips of roughly the same clockspeed. Moreover, it's the only desktop chip currently on the market to make use of DDR3's mahoosive memory bandwidth.

DDR3 paired with an i7 can read/write around 13Gb/s, compared to the around 7Gb/s you'll get from the same sticks with even the most powerful of Core 2 contenders. It's the processor to have, without a shadow of a doubt. Except… there are three big buts and we cannot lie.

First off, the chips sure ain't cheap. You're looking at just short of £300 for the entry-level i7, and around £850 for the biggest daddy. Second, it uses a new socket, LGA1366, so you'll definitely need a new motherboard as well. Thirdly, those mobos sure ain't cheap either. Good luck finding a 1366 board for much less than £200

• Read: Fast motherboards for Core i7

So, a massive outlay buys you the peace of mind of future-proofing and an incredi-PC. Well, in theory.

Bear in mind that the i7 and DDR3 doublewhammy won't make an enormous difference to most games, since few have decent multi-threading support. You'll feel the boost in video encoding and high-end Photoshop work, but if you're strictly a devourer of entertainment then right now a full-on i7 setup doesn't get you much more than willy-waving points.

Well, unless you're using multiple recent GPUs – for instance, SLI or Crossfire X – in which case the gains get a lot better, because your PC's performance won't be bottlenecked by the graphics card. Of course, if you're the type of nutter who already has three high-end GPUs, then no doubt you're rich and crazy enough to pick up an i7 setup without worrying.

It is a recurring theme in modern motherboards that old parallel buses are being replaced by new, shiny, high-speed serial ones. The only major exception to this rule being the memory bus: see Memory Tech opposite for more details on that. Other than speed – reason enough for users to love serials – from a manufacturing point of view, it simplifies design and cuts costs, as fewer numbers of lines are easier to route around the motherboard and result in less layers being required, significantly reducing costs.

The most obvious addition to the serial party was the introduction of PCI Express. This usurped PCI and AGP as the primary way to add graphics and expansion cards to a PC. Attached to the South or Northbridge, PCIe slots have multiple lanes associated with them – 16 for a graphics slot offering 4GB/s worth of bandwidth. The new PCIe v2 standard doubles this rate and is supported by all the latest chipsets.

Multiple-GPUs

The biggest side-effect of PCIe was the ability for motherboards to suddenly support multiple graphics cards. As you probably know, Nvidia offers SLI and AMD CrossFire – the latter also supported on Intel's high-end X chipsets – for multiple GPU support. Whilst twin 16-lane PCIe slots have been the norm until now, expect new boards to offer three slots, as we move into a period where GPUs will be used for physics acceleration. So it's a case of more the merrier… Oh, our poor steaming power supplies!

Of course, it's one thing to know the theory behind how new technologies can be connected together on a motherboard, but it's another to know where you're going to physically pop your processor, memory and expansion cards. That's handled by the ATX specification, now up to version 2.2.

Despite being incredibly dull, it's vitally important and specifies crucial things such as: what size the motherboard can be, where mounting holes are located, power connections, power levels, physical positioning of components, colour of connections and so on. Boring maybe, but if vendors decided they'd put these things wherever they liked, it'd be chaos.

ATX still king

Despite being around since 1996, ATX has staved off multiple attempted coups, including BTX, introduced by Intel in 2003 mostly as a result of the soaring power output of the old Pentium 4. It offered optimised airflow, a larger heatsink and motherboard area with dedicated low- profile designs. But as the P4 faded, any justification for the increased costs of BTX became extinct as well.

We're seeing ever-increasing drives to lower power consumption and increase efficiency: laptop mobos use technology that can reduce idle Northbridge power to below 1W, though an entire running desktop system board will suck at least 40W idle, compared to the 10-20W for an entire laptop. The VIA mini-desktop ITX boards strike a balance on this front, usually consuming 10-15W in idle and usually less than 20W under full load, but that's hardly ever going to be a realistic choice for the gamer or anyone else that wants to perform other intensive tasks.

Part of the problem is just power regulation, that's pulling the 12v line down to around 1.5v for the processor. Much like PSU efficiency, motherboards have the same problem. Much noise is made about multi-phase voltage regulators: while we're sure they do smooth the voltage in technical terms, they seem to have little real effect. Similarly, new energy regulation modules such as the Gigabyte DES or Asus EPU claim to regulate voltages and perhaps clock speeds more efficiently. These do affect power consumption, perhaps as much as 10w or more at full load.

Despite some shrinkage in the market over the last few years – VIA being sidelined, ULi being bought by Nvidia and names like ALi and SiS slipping into the background – there's still a lot of choice if you're after a new mobo. Higher levels of integration mean you get more for your money; budget boards come packed with features while today's high-end boards come bristling with exotic extras, to the point of confusion and pointlessness.

Intel also on top

Intel remains the leading chipset supplier: about half the world's mobos have "Intel Inside". It's also a technology leader, bringing new developments such as PCIe v2, SATA 2 first to boards in new chipsets. It has used the Memory Controller Hub (MCH) as the Northbridge and the I/O Controller Hub (ICH) for the Southbridge ideology since the debut of the P4. ICH9 is the current mainstream Intel Southbridge and supports up to 12 USB 2.0, six SATA, two eSATA and twin Gigabit network ports, plus high-def audio and six single PCIe lanes. Extra options can supply hardware RAID and Turbo Memory. ICH10 is just becoming available but adds little to the party: you should note there are no PATA channels at all.

Robson technology, now branded Turbo Memory, was supposed to increase drive performance, but real- world results are disappointing. ICH10 introduced User Pinning, so it's possible to permanently store an application in the flash memory. This could have been an interesting development, but it has been somewhat overshadowed by the likes of Asus' Express Gate that enable you to boot into a "Virtual Appliance Environment" (VAE) featuring a basic GUI, a browser based on Firefox 2.0 and a Skype VoIP client in less than six seconds, courtesy of a compact Linux distribution.

This 'instant-on' OS offers many advantages: it's able to run on PCs where the hard disk is absent or malfunctioning, and is less vulnerable to malware as the system is mostly read- only. Already a feature of the Eee Box and nine others, Asus will be extending Express Gate to its whole range in 2009.

Intel's high-end option remains the X range; the X48 is the only motherboard to officially support the 1600FSB for both DDR3 and Core 2 Extreme Edition. In reality though, most available X38 motherboards can actually offer this support anyway, as well as the XMP for enhanced memory performance profiles. It also offers two 16-lane PCIe v2 slots for CrossFire support.

A similar situation can be applied to the new P45 that replaces the P35. At least it does introduce PCIe v2 and optionally supports three graphics slots, but as it's positioned as a mainstream chipset, it's hardly required at this point. If your modules support it, Intel XMP memory technology will mean you get the best performance out of them.

What about Nvidia?

Nvidia has always made a point of offering both Intel and AMD options. Currently, its 790i Ultra SLI with its top-end Intel chipset, boasts: PCIe v2, support on two full 16-lane graphics slots and a third PCIe v1 16-lane slot, DDR3 up to 2000MHz and 1600FSB support. The accompanying MCP Southbridge offers an additional 12 lanes of PCIe, twin Gigabit ports, ten USB ports, high-def audio, six SATA and two PATA ports and even a PCI bus.

Unlike the Intel chipsets, Nvidia has opted for DDR3 only, firmly encamping at the high-end. It of course offers the NVIDIA "SLI-Ready Memory" technology for EPP memory support. For people requiring DDR2 support, the 780i still offers triple SLI in a largely similar configuration to the 790i and the 750i with its 'vanilla' PCIe v2 SLI support.

With its AMD range of chipsets Nvidia's entire range sports integrated graphics – DirectX 10, HDMI and HDCP- compatibility. We guess the lack of a memory controller on the Northbridge helps when things get busy, almost like having a spare room in the house. The 780a is its high-end option and supports three-way SLI – again, via two PCIe v2 and a single v1 16-lane slots.

Oddly, the SLI support hangs off the Southbridge, that being its sole job. The usual Southbridge connections are on the Northbridge, offering 12 USB 2, six SATA, two PATA and Gigabit port, plus high-def audio and three more PCIe v1 lanes.

For mid- and entry-level buyers, the Nvidia 8300/8200 chipsets are the perfect choice. Identical in everything but graphics speed, these are single chipset solutions, offering the same abilities as the 780a but with only a single 16-lane PCIe v2 slot. This is somewhat balanced via the Hybrid SLI feature that uses the onboard graphics to accelerate graphics or save power. The Hybrid mode only works with the 8400GS and 8500GT cards, which makes sense for low-end options.

AMD still plugging away

Finally, what about AMD? It's still pushing its high-end 700 chipset, and has rebranded the ageing Xpress 3200 and Xpress 1600 ranges to the 580X and 480X to eliminate any confusion from their Intel- compatible counterparts.

The high-end 790FX (launched at the end of 2007) is still something of a power-house. It supports two physical CPU sockets – rare in real-life – along with 42 PCIe v2 lanes enabling four-way CrossFire fun via two full 16-lane and two 8-lane slots. Oh, and twin Gigabit ports – yes, it's a fairly feature-heavy Northbridge, alright.

Odd, then, that it was typically paired with the comparatively weak SB600 Southbridge, offering only four SATA ports, an extra four PCIe v1 lanes, RAID, two PATA channels and 10 USB 2.0 ports. Happily, this is being superseded by the SB700/750, with six SATA ports, eSATA support, 14 USB ports and support for hybrid fl ash drives.

For more mid-range systems, AMD's 790X supports two physical CrossFire slots and the new 780G. This last chipset is of interest as it comes with integrated graphics and offers similar features to the Nvidia 8300/8200 chipsets. The Hybrid CrossFire works alongside a HD2400 or HD3450 graphics card for about 50 per cent increases in 3D speed.

Those will remain on the market for a good while. New high-end options such as the Intel X58 and the AMD 800 series are on the way with new sockets and better memory support, but not until the beginning of 2009. So why wait?

The complete guide to upgrading your PC

AMD's processors may not have the raw power of Intel's Core i7, but they offer great value for money

According to Microsoft, less than five per cent of end users of its operating systems open up their machines. Given that upgrading is one of the cornerstones of what has traditionally made the PC such an attractive platform, that's an amazing statistic.

Why aren't more of us extending the life of our machines with a simple upgrade? Has our throwaway society really reached the point where it's easier to replace the whole machine rather than keep it up to date by upgrading?

Perhaps people worry that their warranties will be invalidated, or think that swopping components around is more hassle than it's worth. Is Microsoft making upgrading desktop computers tougher than it should be?

That five per cent statistic surfaced in defence to Microsoft's policy on having to reactivate Windows when you change a significant number of components in your machine. Adding more memory or replacing the graphics card shouldn't incur the wrath of Redmond. But touch the motherboard or the main hard drive, and you'll probably need to reach for the phone.

Reactivating Windows is little more than an annoyance, though, so it's hardly a deterrent. Is the real reason simply that upgrading has become too complex and expensive? Over the next few pages we'll look at the many different options available, what tricks you need to bear in mind when physically upgrading and what you need to look for when considering a prospective upgrade.

We'll also look at the performance advantages of some of the key components and the trends and big releases that will change the upgrading landscape in the coming months. Upgrading your machine isn't a dark art. So while delving into the depths of your PC case might sound scary, don't be frightened. Let us take you by the hand and ease you into getting started on making some essential upgrades to your system.

Motherboards

Picking a motherboard isn't like picking any other component in your machine – it's less of a component and more of an assertion of a choice of platform. How you pick a motherboard is more about everything else that's going to end up in your machine rather than which features you would like on it.

Your choice of motherboard can essentially be reduced to asking which of the two CPU manufacturers you want to patronise, as Intel and AMD haven't shared a common platform since the days of the 486. So only once you've picked your processor can you decide which motherboard you want to use in your system.

MOBO CHOICES: Don't spend more than you need to – pick a motherboard that has features that you'll actually use

The latest platforms from Intel and AMD represent a fundamental change in the role of the motherboard. Both Intel's Core i7 and all of AMD's Athlon/Phenom/Phenom II CPUs now feature integrated memory controllers. At the most basic level, this means that the raw performance that used to separate the various motherboard chipsets has vanished.

Nowadays, apart from additional logic and a better selection of voltage regulators and capacitors to assist overclocking the chips, there's little to separate motherboards in terms of data throughput. If you're looking to differentiate motherboards by way of features, you're going to need a magnifying glass, as the most essential features are now included across the board.

This isn't due to a malaise among the motherboard manufacturers, but rather a shift in where the features are integrated. Indeed, it isn't the motherboard manufacturers that offer up the core competencies these days – it's the chipset manufacturers. For instance, you no longer need to go out of your way in order to find a motherboard offering RAID support, because RAID 0/1/5 is now standard on all modern chipsets.

There aren't as many companies manufacturing motherboard chipsets at the moment either – Nvidia and AMD produce chipsets for AMD's CPUs, and Intel covers its own bases. Intel's Core i7, for instance, is currently only served by one motherboard chipset: Intel's own X58. Motherboard manufacturers have managed to tweak this situation to address a number of price points, but these are generally around the £200 mark (with one exception in the form of the Gigabyte GA-EX58-UD3R, which can be picked up for £143).

The feature-sets are impressive, but the lack of a truly affordable motherboard has undoubtedly held the Core i7 back. Gigabyte managed the low price on the GA-EX58-UD3R by shaving off two memory slots, which capped the maximum memory offered by the motherboard – although realistically, four slots should be sufficient for the vast majority of users.

There are still a few differentiating features between motherboards, though, with audio support being a major selling point for most manufacturers. How many Gigabit Ethernet ports are included can also separate one board from the next, while those looking to attach FireWire peripherals will find themselves having to pay out a little more to find a motherboard that supports the technology.

Indeed, for many it may be worth buying a more affordable motherboard and a FireWire PCI card instead of one that offers support natively. A selection of premium motherboards fall under the 'enthusiast' banner, with those featuring overclocking options costing more than normal boards. A card with a back-up BIOS will set you back more, for instance, but it can save you time if you push an overclock too far.

Onboard power and reset switches can also be useful if you like to tinker inside your machine, although again these tend to be available only on the higher end boards. If you're planning on running multiple graphics cards, one of the most impressive solutions currently available is Intel's X58 chipset.

This is the first time that a single motherboard chipset has been able to support both ATI's CrossFire and Nvidia's SLI technologies. We'll cover this in more detail when looking at graphics card upgrades, but for those who like the idea of multi-GPU graphics but don't have a particular allegiance, it certainly makes Core i7 a tempting platform.

Ultimate upgrade: If you're looking to get the full gamut of features, the Gigabyte GA-EX58-Extreme ticks all the right boxes, including FireWire support, HD Audio and up to 24GB of DDR3 RAM.
Price: £231 (£201 ex VAT)
Info: www.giga-byte.co.uk

Best bang for the buck: AMD's Phenom II processors may support DDR3 memory, but the difference over DDR2 is slight, making the likes of the Asus M4A78 a wise buy for those on a budget.
Price: £60 (£52 ex VAT)
Info: uk.asus.com

Hold on for: The release of Core i5 should produce a far more affordable (although annoyingly incompatible) platform compared to Core i7. Expect to see boards costing closer to the £100 to £150 mark.

CPUs and RAM

CPU (Intel)

The launch of Intel's Core i7 has changed the CPU landscape considerably. Intel already had a lead on its old rival AMD thanks to the raw power of the Core 2 Duo. While these chips were somewhat inelegant, they delivered where it counted most: raw performance.

CORE I7: These are a good choice if you need a lot of power but you might do better waiting for Core i5

Core i7, however, has opened a gulf between the two companies and, thanks to the inclusion of a memory controller, meant that AMD's often-lauded memory throughput has finally been eclipsed. Intel has recently released new revisions of the Core i7, and the D0 stepping of the Core i7 920 has quickly become the most sought-after processors around, thanks in no small part to its overclocking potential.

However, while the chip itself isn't too expensive, you'll need the whole platform to make use of it, and that can cost a surprising amount of money when it's all added up. You'll need a socket 1366 motherboard to plug your processor into, enough memory to make sure the new processor has room to breathe (6GB has quickly become the norm) and of course the processor itself.

You're quickly looking at £500+ for such an upgrade, which is why AMD has managed to keep competitive.

Ultimate upgrade: The eight logical cores of the Core i7 represent the pinnacle of desktop computing. The new 920 D0 stepping offers stunning value for money, too, especially if you overclock it.
Price: £216 (£188 ex VAT)
Info: www.intel.com

Best bang for the buck: Multicore processors needn't cost the earth. The AMD Phenom II X3 720 Black Edition isn't just an affordable processor, it's also the basis of an affordable platform that can still deliver big numbers.
Price: £109 (£95 ex VAT)
Info: www.amd.com

Hold on for: It's not going to trouble the Core i7, but Intel's Core i5 boasts a more mainstream pricing for the whole platform that could provide a knock-out blow to AMD's current value processors. Expected this Autumn.

RAM

There are four types of RAM currently in circulation, and they're defined by the support offered by your motherboard. The newest addition to the world of memory is DDR3, and it's an integral part of Intel's Core i7 and AMD's AM3 platforms.

However, older DDR2 RAM is still the most common memory standard (Core 2 Duo and Phenom both use it), and for good reason: it's affordable, well supported and has a proven track record. The third most popular type of memory is DDR, which could be the memory in your machine if you haven't upgraded in recent years.

TRIPLE CHANNEL: DDR3 demands you have three sticks of RAM, one for each slot on your mobo

If you've got a very old machine then you may still have basic SDRAM, in which case you'll find upgrading surprisingly expensive as it's now very much a niche technology. Whichever memory standard you're using, RAM sticks are defined by their speed and latency.

There are a few things to watch out for: you need to get the new sticks to match the memory that's already in your machine (or alternatively throw that away and start afresh), and you need to buy in multiple numbers that match the number of memory channels on your motherboard.

So if you're running a Core i7 rig, you'll need three DIMMs, because Core i7 boasts a triple-channel memory controller. If you've got an AM2+ Phenom II, then a pair of sticks is the order of the day.

Ultimate upgrade: You can get far more affordable 6GB kits, but the Corsair Dominator GT set boasts support for 2GHz bus speeds and 8-8-8-24 latencies for an impressive slice of memory performance – and it looks impressive too.
Price: £237 (£206 ex VAT)
Info: www.corsair.com

Best bang for the buck: If money is more of an issue, the OCZ Gold 6GB 10666 triple-channel kit makes for the perfect accompaniment to a Core i7, and with 9-9-9-20 latencies it's not sluggish either. Price: £63 (£55 ex VAT)
Info: www.ocztechnology.com

Hold on for: There are no major new memory technologies on the horizon, and DDR3 memory prices are finally at reasonable levels. Prices should continue to fall a little over time, but until Core i5 arrives and shake things up further, now is a good time to buy.

Graphics cards

Graphics cards

The graphics card is the most likely of all modern PC components to start a heated debate. Why? Because anybody who has bought one of the latest and most likely expensive pixel-pushers will be keen to justify their purchase. However, when all the passions have been spent and the dust has settled, there is little separating the latest offerings from AMD and Nvidia.

As ever, though, the devil is in the detail. A fundamental shift has taken place between AMD and Nvidia's graphics cards recently, and it's to do with how the companies are approaching the highest end of the 3D market.

Nvidia has largely maintained the stance that it can keep improving and tweaking its cores to produce faster and faster GPUs, and to some extent it has succeeded. AMD, on the other hand, has concluded that multiple GPUs are the answer for high end products, and so it has been focusing more on the mainstream market, with high-end solutions made up of combinations of the mainstream chips. And again, the company has been successful to a degree.

What this means to us as upgraders is that there is a wide range of graphics cards available that can transform your gaming experience. Specifically because of the refocusing, an incredible amount of GPU power can be had at the more mainstream end of the scale, with cards costing between £100 and £150 now more than capable of driving a 20in or 22in screen at the native resolution with all the effects turned on.

You'll need a bigger screen in order to get the most from cards that cost close to £200, with the likes of the GeForce GTX 275 capable of running a 24in screen at the highest settings at a smooth 60fps. Indeed, it's only when you connect a machine to a 30in screen that you really push the current generation of cores, and it's then that you either have to select a single-card multi-GPU solution or go for an SLI or CrossFire card to drive the screen properly.

At this point you can easily spend a small fortune on a graphics subsystem (and the accompanying power requirements) for your system, but with good reason: the end results are usually extremely impressive.

An alternative way of using SLI and CrossFire is as an upgrade path in itself. Buy a card now, and then buy a second card later for a moderate performance boost. It's not a bad theory, but in practice finding a card a couple of years out of date is tricky, and stockists tend to charge roughly the same for older cards as they do for a more modern card.

The second-hand market can help you track down that second card, but there will be no reassurances that the card will work, so it can be risky. You're looking at roughly a 60 to 70 per cent performance improvement from the addition of that second card, so it may not transform your gaming the way that you were looking for anyway. And while driver support for such solutions has improved, it's still not perfect.

Essentially, we'd recommend caution when it comes to both SLI and CrossFire graphics cards, unless the card in question is so affordable that it becomes a risk worth taking. One thing you do need to be aware of when eyeing up a potential graphics card is that you'll not only need enough power to run the card, you'll also need the connectors to do so.

The more mainstream cards require just a single six-pin connector, but faster cards will require either a pair of these, or one six- and one eight-pin. You need to make sure that there's ample cooling in your machine too, as some of the cards (particularly dual-GPU solutions) can reach close to 100°C under full load.

Once you've decided what sort of price you can afford and selected the GPU you want driving your graphics card, things come down to the clockspeed of the GPU and RAM and the amount of RAM on each card. There's an economy of scale here, too – big manufacturers can afford to produce graphics cards that little bit cheaper. Keep an eye on recommendations in PC Plus for the latest and greatest cards that you should be considering for your rig.

Ultimate upgrade: If you're looking to drive a large screen, the Asus GTX 295 is an incredible piece of graphics kit. Capable of running all modern games smoothly, it's a stunning one-stop card.
Price: £360 (£313 ex VAT)
Info: uk.asus.com

Best bang for the buck: It's a close battle between AMD and Nvidia for best value, but the recent price battle makes the Sapphire HD 4870 1GB the best choice.
Price: £132 (£115 ex VAT)
Info: www.sapphiretech.com

Hold on for: Things should hot up once Intel enters the market with Larabee, and assuming that it's any good, we should see the price battle intensify even further. On paper it looks very interesting, so if you fancy giving it a try, hold your horses on the graphics card front for a bit.

Hard disks and power supplies

Hard disks

Hard drives have maintained that unerring pace of progress that only computer components seem to achieve: the core technologies behind them are continuing to improve (leading to larger capacities and faster throughput), while the price tags just keep on dropping.

As a snapshot of the current state of the hard drive market, you can pick up a 1TB hard drive for as little as £60 to £70. The technological advances mean that such drives perform well too, with some managing read speeds as high as 80MB/s. If you want serious performance, though, there is a slightly more newsworthy addition to the storage market in the form of 'affordable' solid state drives (SSDs).

OLD FAITHFUL: Although SSD's offer much faster speeds, for cost you can't beat the humble harddrive

This technology has been rumbling along for years now, but the launch of the Intel X-25 pushed SSDs out from their niche to become a serious consideration for anyone looking for highperformance components. High-end SSDs offer read speeds of over 220MB/s, and much improved write performance.

Even if you can't afford a top-of-the range SSD, a low-end drive will generally give twice the performance of conventional platter-based drives. Ideally you want to set the SSD up as your main boot device, as it can as much as halve your boot time.

There are two significant downsides to SSDs, though: their capacities are tiny when compared to platter based drives, so you'll need one of those for your data anyway, and they're significantly more expensive than standard hard drives. If you can afford it, the best solution is to use an SSD as a boot drive and a conventional platter-based drive for your data.

Ultimate upgrade: It's not cheap and it's not particularly big, but it's incredibly fast – Intel's X25-E has made SSDs a real consideration for those looking for performance. It may hold only 32GB, but the X25-E will leave your existing hard drive looking pathetically slow.
Price: £325 (£283 ex VAT)
Info: www.intel.com

Best bang for the buck: The perfect accompaniment to an SSD or indeed as a main drive for the more cost conscious upgrader, the Hitachi Deskstar boasts a 16MB cache, 7,200rpm spindle speed and a cool 1TB of storage for a paltry sum.
Price: £59 (£51 ex VAT)
Info: www.hitachigst.com

Hold on for: The SSD market is still growing, but expect prices to settle a little over the coming year, thanks in part to improved support from Windows 7. Hybrid drives may off er the best of both worlds, so keep an eye out for them.

Power supplies

While it would be reassuring to think that the massive success of netbooks means that low-powered computing is infiltrating every aspect of our computing lives, the truth is that performance computing continues to increase the demand placed on our power supplies.

While we used to get by with the 250W PSUs that came bundled with PC cases, if you mean business then you're looking at making a separate purchase these days, and one that can set you back a tidy sum.

BFG: Remember to leave a little bit extra power with your PSU in case you need to upgrade again

It's not all bad news for the environment, though, as manufacturers are taking efficiency seriously – a lot of PSUs already offer 80 per cent efficiencies, but 85 per cent and 90 per cent models are on the way too. The hardest part about picking a power supply is working out how much power your machine needs.

Thankfully there are freely available tools to help you work this out, ranging from the simple to the exhaustive. Once you know how much power your system (or your potentially upgraded system) needs, give yourself a little room for further upgrades (50 to 100W) and you're done.

Ultimate upgrade: It's undoubtedly overkill for the vast majority of machines, but at least with the CoolerMaster Real Power 1,250W PSU you know that you won't be caught short on the power front, even if you fill your machine with all the best kit.
Price: £150 (£130 ex VAT)
Info: www.coolermaster.com

Best bang for the buck: Realistically your PSU needs are going to be quite modest, and a 500W model should really be more than enough. If you don't fancy shelling out too much dosh, the OCZ StealthXStream 500W Silent should have you covered for an upgrade or two.
Price: £43 (£37 ex VAT)
Info: www.ocztechnology.com

Hold on for: More efficient PSUs that can hit 90 per cent efficiency at high load are on their way, and while they may be more expensive to buy initially, they'll save you money over time.

One essential upgrade...

Sometimes you just can't afford to upgrade your whole system. If you can afford one upgrade, and only one, then how you spend your money should be dependent on what you use your machine for. Gamers have it easy – a new graphics card is generally the way to go as long as the processor isn't more than a couple of generations old.

If you move a lot of data around then a new hard drive makes sense. For heavy application use, the difference having an SSD as the main drive can make is simply incredible. However, for general day-to-day use, and for a more rounded upgrade for your money, we'd have to cheat a little and recommend that the move to the new Core i7 platform is the best upgrade you can make.

Intel has done a sterling job of shattering many of the bottlenecks of previous generations: the memory throughput is stunning, and the raw power of Core i7 makes light work of traditionally intensive tasks. It's not without its downsides, of course – and in the case of Core i7 the downside is very much financial – but if you need the power, there's nothing quite like it.

At least for now. It could easily turn out that the biggest threat to Core i7 is its soon-to-be-released sibling, Core i5.

How motherboards are made

We take a look around Gigabyte's production facility

<>

3 million motherboards is a lot. Just think about the amount of solder alone that must take, the amount of Japanese capacitors, the extra amount of copper needed to fulfil Gigabyte's 2oz copper pledge for each of its mobos.

The thing that will always stick with me from the beating heart of Gigabyte's Nan-Ping factory is the incredible amount of human labour it takes to assemble and test each and every board. That's the amount of boards the factory is capable of producing every year; three-million boards. In a single year. Assembled and tested by hand.

But the manufacturing process, with all its immense scale, is only part of story. Before the Chinese-manufactured slivers of PCB even make it to the gates of Gigabyte's high-tech factory on the edge of Taipei's Xindian City, the design process will have been in full swing, working on a new chipset design for a full year, while the research involved stretches back even further.

Sitting among the clouds above Taipei's banking district, high up in the 101 Tower, I've been talking to Gigabyte VP of Motherboards, Henry Kao and manager of the Product Planning division, Jackson Hsu.

With Intel's P55 chipset all set to go global, we've been chatting about what exactly goes into putting a new chipset design out on the street.

Self-referential

The Taiwan-based company has been working on the latest P55 chipset for the last year, as one of Intel's key development partners.

"We don't need the Intel reference design board, because we make the reference design board for Intel," says Henry Kao grinning broadly. "Intel has the idea for a new chipset and once they get it to a certain level they bring the design idea to Gigabyte so we can co-design the reference board."

It's this board that gets taken around the other labs so Intel can show them the new design layout. "With a new chipset we usually spend one year co-developing with Intel. Once it feels comfortable then it releases that design to the other motherboard makers. From there we'll make our own specifications and it'll be another month before the first working samples are available," says Henry Kao.

The initial design teams working on the co-development project will usually be a very small group of people. "At that stage you don't need too many people involved, just a single team, dedicated to Intel," says Jackson Hsu.

"Once the chipset is more mature, then we expand the team." Once the reference design has been finalised this is where the motherboard manufacturers can begin to create their own different spins of the same board to create their high, mid and low-end iterations of the chipset.

As a manager in the product planning division it's this part that Hsu has to be heavily involved with: "As a planner you know the basic chipset and CPU layout but need to start thinking what do I add? What would my competitors do and what does the customer need? And what is the latest technology? Then you have all the elements you need and you have to start picking them out. What do you need for your highest high-end, then lower down the pricescale, what do you take off? It's the same process every year, but the elements are different."

So how do you choose which components go into the higher-end boards and which get taken out? As ever the over-riding principle is based around cost. Sticking in the extra couple of USB slots that Intel are asking for on the P55 boards isn't a problem across a full range of mobos due to the miniscule cost, but other parts inevitably need trimming off the high-end boards.

Counting costs

Where the lower end motherboards suffer then is in the newer technology: "I can't have everything twenty-four phase," says Hsu about the twenty-four phase power that's used for stabilising the power signal of the top-end P55 motherboards.

That has to be reduced further down the board pecking order from twenty-four, to twelve to eight: "It's also about 6GB/s SATA we're using this year. Most of our P55 boards will have 6GB/s SATA; on the mainstream and high-end we'll have four ports, but lower down the line we'll have two ports. So even at the low-end consumers will still have access to 6GB/s SATA, but only two ports."

But cost isn't necessarily the over-riding principle governing Gigabyte's motherboard division as a whole. The current division CEO is a certain Mr. Lin, a man who has risen through the ranks right from the engineering floor. Now there's a man at the top who has hands-on experience of what goes into both the design and manufacturing processes.

This is where the aforementioned 2oz of copper came from. Slotted between the layers of PCB is this layer of copper, absorbing and distributing the huge amount of heat that is generated by both the processor, chipset and power phase regulators.

It's made the motherboards slightly more expensive to produce, but in the long run it has reduced the number of RMAs to such an extent that all of Gigabyte's motherboards are now adopting the same process.

Moving half of the company's motherboard manufacturing to this fairly new Taiwan-based factory isn't about cost either. The factory workers expect, and get, a far higher wage than their Chinese counterparts in Shenzhen, for example. And with three shifts working eight hours each that's a lot of man (or young girl for the most part) power to pay for every day.

But this factory is catering for the high-end motherboards and graphics cards, as well as notebooks, mobile phones, servers and full desktop PCs. And having a hi-tech factory of this sort on the doorstep of the research and development teams at Gigabyte means that it can rapidly turn around any design changes that it needs to make. And as Jackson Hsu told me later: "Some of our customers still like to have that 'Made in Taiwan' stamp on Gigabyte products."

Making the motherboards

So once all the design work has been done, and all the various iterations of a specific chipset have been thrashed out, it's on to the actual manufacturing process itself. The first thing that hit me – after I'd been kitted out in sexeh blue overshoes, swine-flu mask and had every mote of dust blown off my bloated form in the biggest hairdryer this side of Tony and Guy's HQ – is the amount of motivational material dotted around the factory floor.

Much of it seems to be for the international visitor as it's all in fairly broken English. In particular was a Deirdre's Photo Casebook-style storyboard showing how a poorly fixed nut caused the explosion of a Chinese Boeing 737 - a lesson to focus on even the smallest things or face the direst of consequences.

Another favourite was a sign in both Chinese and English instructing the workers to 'Be more responsible, complain less, be more attentive and make lesser mistakes.' It all seems to work though as the Nan-Ping factory is capable of shunting out a quarter of a million motherboards every month, as one of four Gigabyte factories in Taiwan and mainland China.

With a production facility working literally around the clock, across a range hardware, including graphics cards and servers as well – producing 50,000 and 5,000 per month respectively – the manufacturing bunnies have got their work cut out for them. There are four stages to both the motherboard and VGA manufacturing process carried out in Nan-Ping.

First is the Surface Mount Technology (SMT), where the smallest components are placed on the bare printed circuit boards (PCB). Second is the Dual Inline Package (DIP) where the larger elements are attached, then there's the testing stage and then finally it's down to packaging.

Each of these processes is carried out on separate production lines, and often on separate floors. For a single motherboard to go through the entire process though takes only around fifteen minutes from the start of production to boxed finish. The first step, the SMT process, is the most automated of the lot.

All of Gigabyte's PCBs are manufactured in mainland China to its own specifications and shipped over, bare, to the various factories. Once they hit one of the eleven SMT production lines in the Nan-Ping factory, each PC board is put first through a solder paste printing machine. This machine lays out the areas of solder needed to hold the components, which are then placed in the following step.

The smallest components, right up to the northbridge and southbridge chips, are held in different sizes of rolls, feeding their chips into either the high-speed or multi-function placers. These are like component nail-guns, shooting chips, resistors and capacitors into the motherboard with almost reckless abandon. Actually, they are in fact incredibly precise, with the high-speed placer inserting a component every 0.1 seconds.

Once all of these smaller parts have been placed on the board the entire PCB gets moved into an oven for what is called reflow soldering, which is basically melting and re-setting the existing solder to lock the new components in place.

The final part of the SMT process is the testing. At each step every motherboard and graphics board has to go through the same testing procedure; first comes the inspection, then in-circuit testing (ICT).

In this case, with up to 1,300 components on a motherboard, the inspection is done automatically with an optical inspector, which uses a strobe to check each component is correctly placed. In seconds. Then there's the first manual inspection, then the in-circuit testing which checks the electrical performance of the newly-placed components.

Dip and cover

This is where things become a whole lot more labour intensive; this is the DIP process. There are only four production lines in this part of the factory compared to the eleven of the SMT, but there are fifty operators on each of them manually inserting parts into the PCB.

The components that are added in this section are things like the PCI connectors, memory slots and all the inputs and outputs. The board is then put through a wave soldering machine, once all these larger components are in place to secure these latest additions and then passed onto another manual technician to touch up any missing solder or unsecured components.

It's at this stage that all the heatsinks are attached to the chipset. Once attached, there follows an almost identical inspection and in-circuit test to the SMT process. On the same floor as the DIP procedure are the six testing lines. This is where all of the boards, motherboard and VGA, get a complete check up.

First is the function test, where the motherboards are placed in testing benches and get a full functionality check-up. Then comes the burn-in test where each board gets subjected to a high of 45°C and then a low of -10°C, just to make sure the final setup can cope with the daily rigours that it might come into contact with.

From here it's on to the final manual inspection and then on to that final stage of packaging then on to retail. With such finely tuned electronics, it's always tempting to think of it being a completely automated process, something akin to a car manufacturing plant but with far, far smaller mechanical arms. The first time I had a chance to tour such a factory though it surprised me about the sheer level of human labour involved in everything.

So when you next come to buy a motherboard or graphics card, or even just take a peek inside your PC, take some time to think of exactly how many hands that bit of PCB has gone through.

Nvidia GeForce GTX 200 vs. ATi Radeon HD 4800

Size isn't everything. The latest offerings from the two big players look very different, but the results are less clear cut

About three years ago, PC processors went through a technical – and somewhat philosophical – revolution.

The Intel Pentium D of 2005 was the first dual-core CPU and it signalled the end of the quest for ever higher clock speeds. Instead, the CPU industry shifted to a theoretically more efficient multi-core, parallel processing approach.

The first signs of an equivalent change in the graphics chip market appeared in 2007. First, Nvidia elected not to replace the GeForce 8800 GTX GPU with a large, significantly more powerful chip. Its alternative was a die-shrunk GeForce 8800, focusing on efficiency and cost savings rather than outright performance.

One step beyond

AMD's graphics subsidiary ATi took the same approach when it moved from Radeon HD 2900 to Radeon HD 3800 Series GPUs. In fact, it went one step further. ATi released the dual-chip Radeon HD 3870 X2 and announced that as far as it was concerned, the game was up for big graphics boards based on a single monolithic GPU die; the future would be multi-chip.

All of which sets up a rather intriguing backdrop for the introduction of a pair of new GPUs from ATi and Nvidia. Once again, ATi has focused on efficiency and affordability.

Nvidia, on the other hand, has gone old school and delivered a single-die graphics chip of truly mammoth proportions.

Nvidia's attempt

You might think that these new pixel pushers are not directly comparable, but the competition between ATi and Nvidia will be just as fierce. The difference now is that the contest is no longer a clean fight between two graphics chips. Instead, it's a battle between two distinct design philosophies and business models.

First out of the blocks is Nvidia's beastly new GeForce GTX 200 series. By any metric, it's a monumentally powerful – even intimidating – new graphics chipset. The GPU at its heart contains an incredible 1.4 billion transistors. That's literally double the number in GeForce 8800 series GPUs. Consequently, the GTX 200 series packs some seriously beefy specifications.

In terms of functional units, the shader count is up from 128 in the GeForce 8800 series to 240. Things get a little more complicated when it comes to comparing new and old texture processing and pixel outputs per clock. Suffice to say that with 32 render output units and 80 texture address and filter units, Nvidia has boosted the new GPU's functional heft in all parts of its architecture by at least 25 per cent – and usually much more.

Long memory

The final piece of the GTX 200 puzzle is memory technology. Here Nvidia has also taken the big iron approach by pairing established GDDR3 memory with a beefy 512-bit interface.

All told, the GTX 200 is a massive 576mm square chip. It's so big that no more than 100 can be squeezed onto the 65nm wafers used by Nvidia's production partner TSMC. To put that into context, Intel can cram approximately six times as many dual-core Penryn CPU dies or 25 times as many Atom processors into the same space. In other words, Nvidia's new GPU is an extremely expensive chip to manufacture.

Unsurprisingly, it's also extremely expensive to buy. At launch, two models are available, the GeForce GTX 280 and the GeForce GTX 260. The former is the full-fat offering with all of the abilities detailed above, core and shader clocks of 602MHz and 1,296MHz respectively and a memory frequency of 2.2GHz. It's yours for around £400.

The 260 model is comparatively cut down with 192 shaders, 64 texture address and filter units, and 28 render outputs. Operating frequencies are likewise somewhat suppressed at 576MHz, 1,242MHz and 2GHz for core, shader and memory respectively. The 260 must also make do with a 448-bit memory bus. The starting price for this slightly more modest GTX 200 is £250.

ATi's option

The GTX 200 series are big boards with suitably bulky price tags. At first glance, then, it certainly looks like ATi has opted not to compete.

Its new Radeon HD 4800 series is a more restrained effort weighing in at just 956 million transistors and 260mm squared, the latter figure aided by the use of a slightly finer 55nm production process. It's an altogether smaller, less costly GPU to manufacture.

Initially, there are two boards in ATi's new line up, the 4850 (read our review here) and 4870 (review here). Both share the same number of functional units and are specified with 512MB of graphics memory. The key differentiators are clock speeds.

Fast clocks

The 4850 runs a core clock of 625MHz and a memory frequency of 2GHz while the 4870 ups the ante to 725MHz and 3.6GHz. The 4870's startlingly fast memory speed is courtesy of the first ever use of GDDR5 memory on a consumer graphics card. As for pricing, the 4850 model starts at £125 while the 4870 is a £200 board.

Here's the twist, though. Despite the smaller die size, transistor count and pricing, the new 4800 family comes awfully close to Nvidia's big beast in terms of raw capability. In fact, it actually outpoints the GTX 280 for pure computational grunt, if not 3D rendering throughput. The top spec Radeon HD 4870 board is rated at a monstrous 1.2TFLOPs where the GTX 280 manages just 933GFLOPs.

The secret

To understand how ATi's new GPU pulls that trick off, consider the following facts.

Compared with the existing Radeon HD 3800 series, the 4800 family has nearly triple the number of shaders (800 to be precise, but note that ATi and Nvidia's shaders are not directly comparable) and texture units. And yet its transistor count has risen by just 44 per cent. In other words, what AMD has done with the 4800 series is far more sophisticated than creating an oversized 3800 core.

Every aspect of the chip's architecture has been overhauled with a view not just to performance but also to efficiency. That includes the welcome re-introduction of standard box-filter algorithms for anti-aliasing.

With the launch of the Radeon HD 2900 series early in 2007, AMD dabbled with a new 'adaptive' approach towards smoothing the jagged edges of rendered objects. In theory, it was more sophisticated. In practice, it was just plain slow.

Green flagship

A further benefit of ATi's focus on efficiency is power consumption. Nvidia's GTX 280 guzzles up to 236W, while the flagship Radeon HD 4870 consumes up to just 160W. Also, of the two big players in PC graphics, only ATi's GPUs have support for the latest 10.1 revision of Microsoft's all-powerful DirectX multimedia API.

Until ATi releases the upcoming dual-chip version of the HD 4870, we won't know exactly how this latest round of the ATi versus Nvidia battle is going to play out. But after nearly two years of Nvidia dominance, things are already looking much more promising in the ATi camp.

Review: the fastest graphics card on earth

The ATI Radeon 4870 X2 is effectively two 4870 standalone cards strapped together

We're not big fans of multi-GPU graphics technology here on ye olde TechRadar. Whether you're talking SLI from NVIDIA or AMD's Crossfire, multi-GPU graphics solutions promise much but typically conspire to disappoint.

While they're often the most powerful cards on the market, driver support has left a lot to be desired. And so just when you're booting up Crysis in Extra Extreme Double Ultimate mode, it'll crash your machine and leave you to reboot.

If NVIDIA or ATI could solve this driver support problem, they'd sell a lot more of their dual-GPU cards.

The ATI Radeon 4870 X2 (formerly known as R700) then, has a lot of potential. The 4870 standalone card is the second most powerful single-GPU card on the market - and so this dual card is by far the most potent card out there.

It's got a theoretical maximum compute performance of 2.4TFLOPs. That's the same as the fastest supercomputer in the world from as recently as 1999.

So, has ATI provided us with a beasty card and the drivers to support it amply?

Find out in our full review

Gigabyte Nvidia GTS 250 review

Much has been made of Nvidia's decision to release the GTS 250, mostly down to the fact that the 512MB parts are identical to the practically defunct 9800 GTX, itself a limited reworking of the now-ancient 8800GTX.

Now, we've been praying for the green side of the graphics sphere to simplify its naming structure and with this re-branding of the G92 GPU it's done it, so this is obviously a good thing. Right?

Unfortunately there simply isn't a place for this card right now. With the recent price-cut of the GTX 260, an eminently excellent card, the two stand disturbingly close in the wallet stakes, but a country mile apart in terms of performance.

The naming culture may have been simplified but the pricing structure perpetuates the lie that the GTS 250 offers a viable alternative to its stable-mate.

It's not necessarily Nvidia's fault though, I don't think even they expected AMD to make such a recovery. Certainly not a strong enough one to force the price cuts that have seen the GTX 260 drop well below the £200 mark, squeezing the already paper thin margins for the distributors.

Gigabyte's version though is simply a re-badged 9800 GTX. This 1GB, overclocked offering doesn't carry the die-shrunk G92b GPU, but still sits on a redesigned PCB.

PCB redesign

Gigabyte uses its motherboard tech on the board design to increase voltage stability, has added a tweaked cooler to help drop the GPU temperature further past the reference design, and clocked it a fairly miserly 63MHz above the original 9800 GTX, but with headroom to spare.

Unfortunately none of these tweaks can help it escape the fact that an extra £30 nets you a card that severely outperforms it. Essentially the GTS 250 is running on years-old technology and in some benchmarks, it offers half the performance of the GTX 260.

That said, the GTS 250 is positioned to be in direct competition with AMD's 4850. At the lower end of the resolution spectrum the 4850 stands tall, but as soon as you crank it up, the extra memory of the GTS 250 starts to come into its own.

It's no real surprise then that Nvidia has entered into this re-branding phase – it already had a card to compete with the mid-range of AMD boards, though inevitably consumers will go for the newer release every time. Sound thinking from a business side, and possibly vital for Nvidia's continued development.

Still, what it has over the 4850 is at high resolutions and if you've got the sort of a monitor to make the most of it, realistically you'll save up for an extra week and get a GTX 260 instead. At least, that's what we'd do.

Top 11 budget graphics cards

This week saw the launch of the latest graphics card to bare the mantle 'fastest in the world'. Yep, the ATI Radeon HD 5870 is truly an astonishing card.

The problem is that it costs over £300. And at a time when spending £300 on anything at all is a big deal, let alone a monsterous GPU that no one really needs, it just seems a bit pointless.

Strange as it may now seem then, there was once a time not so long ago when the launch of a new high-end graphics chip was an event we genuinely looked forward to with excitement.

Weeks of rumours, plotting and early silicon sampling would precede a frantic day at launch, when competitor publications and sites were furiously scanned to see how they compared and make sure there wasn't a single detail that we'd missed.

Then AMD bought ATI and decided pixels were no longer as important as price, and refused to compete for top frame rates.

What seemed like an admission of failure to produce a competitive part – and the HD3000 series was pretty terrible – now looks like an act of supreme foresight.

A new direction

For the vast majority of gamers, there's absolutely no need to spend £400 on a graphics card any more.

Here we now have a situation where a whole generation of technical trending has been reversed. Screens are, thanks to the move to 16:9 HDTV ratios like 1,920 x 1,080, getting lower in resolution, not higher.

Games engines aren't getting more complex, because developers are looking at multi-platform releases and concentrating improvements more on the non-graphical elements.

It also goes without saying that we're in the middle of a global recession and blowing your savings on a new GPU suddenly looks a little bit silly too. In short, Nvidia may traditionally have had the very fastest hardware, but it looks a little like an Olympic hero whose homecoming has been ruined by the fact everyone in town is queuing up for the opening of a new Lidl.

Where has the graphics glamour gone?

Feature focus

Vendors are more interested in talking about GPGPU features than trouncing each other with performance. It's almost enough to make you wish for the days of dodgy drivers and bent benchmarks again.

Maybe, just maybe, we've got our priorities wrong.

Instead of stifling a yawn every time another card drops into this already crowded price point and puts out enough pixel power to render a life-sized model of the Burj Dubai for less money than a good meal for four, we should celebrate the ridiculous amounts of processing power on offer, and experiment with how far we can push them – since they're damned near disposable anyway.

Of course, just because the graphics world isn't what it used to be doesn't mean you should just throw your cash away without consideration. There's a staggering amount of choice available for less than £160, and not every card in the category is created equal. Some, in fact are real stinkers.

ASUS EN9600GT - £59

Asus EN9600GT - £59

A stock interpretation of NVIDIA's GeForce 9600GT, Asus' card distinguishes itself in two ways.

First of all, it's the cheapest on test here, which isn't bad for a chip which would have been considered 'mainstream' just a generation ago.

Its second claim to fame, though, is that it's also the worst performing by some way.

If it were a horse race, the 9600GT is the three legged donkey clearly on its way to the knackers' yard. The reason it lags compared to the rest though isn't hard to fathom.

For just £10 more you can bag the far superior Radeon HD4770 (read on), which is a whole class above this for the same sort of money. It's time for the 9600GT to retire gracefully from the field.

Gigabyte 9800GT - £99

Gigabyte 9800GT - £99

Not so long ago, Nvidia's GeForce 8800GT was our favourite graphics chip of the lot. In the same league as the Nvidia GeForce 3Ti and ATI's 1900XT, it had the sort of performance per pound ratio that flew off our charts and shamed all its contemporaries.

Then it was renamed the '9800GT', and we still liked it though – just not as much as before. Hoping to win back some of our initial love for the G92 core, Gigabyte has taken the relatively old chip and invigorated it by fitting an enormous passive heatsink.

Given that game engines really haven't advanced much since the original was pumping out frame rates with the best of them, it follows that decent performance plus absolute silence should be something of a winner.

Not just for a lounge-based media centre, but any gaming rig that you don't necessarily want to sound like a jet engine taking off.

At lower settings the AMD card slaughters it, but again the 9800GT remains playable. So if it's silence you crave and you don't mind paying a premium, it's not a terrible choice.

Read: full Gigabyte 9800GT review

ASUS EAH 4770 TOP - £70

ASUS EAH 4770 TOP - £70

AMD's curveball card really does present a problem for, well, almost everything else out there.

The chip itself is very similar to the much loved Radeon HD4850. What it lacks in physical shader cores, 160 of them to be precise, it makes up for in higher clockspeeds, a fact made possible because it's produced on a mere 40nm process.

It's also tiny, one of the smallest printed circuit boards we looked at, and has an equally small and quiet heatsink. To top off these astounding details, this card also draws around 40W less than the Radeon HD4850, so it's a lowpower card too.

Benchmark-wise, it gives its stablemate a good run for its money at lower resolutions and image quality settings, even beating it on occasion thanks to that higher clockspeed.

The trouble is that prices have fallen so far that you're not actually saving much money by choosing this over its more consistent older sibling any more. A ten percent price advantage might sound like a lot, but since that's only £7 it's fairly negligible in this particular case.

Read: Asus EAH 4770 TOP review

HIS Radeon HD 4830 - £75

HIS Radeon HD 4830 - £75

Looks can be deceiving: take this curious little card which, although more expensive and boasting a more impressive code number than the HD 4770, has considerably less impressive specs.

Introduced at the tail end of last year as a cut down HD 4850 to hit a £130 price point, in many ways it's a surprise to find it still around now that the more talented HD4770 is here.

Featuring the same number of shaders, ROPs and all, the HD 4830 is at a deficit of 225MHz in core clockspeed alone. How can it possibly compete? Perhaps it exists purely to catch the unwary who think the '8' in its designation actually means something?

The card was clearly a stopgap measure which shouldn't have hung around this long. The fact there are so many on sale still gives us hope that the public wised up to this fact early on.

Read: HIS Radeon HD 4830 review

Sapphire HD 4850 512MB - £77

Sapphire HD 4850 512MB - £77

Long a byword for elegance and excellence in the graphics world, this slim single-slot card has topped the reviewers' awards since it launched over a year ago.

And that makes it almost 101 in videocard years yet still going strong. The reason is simple, it may not top the performance charts in many tests, but it's good enough to game on if you're using anything smaller than a 30-inch monitor.

And it runs so quietly that you'll hardly know its there. It's also improved with age. Driver revisions have put it well above the 9800GT with which it once competed in terms of frame rates, and time has brought the price down to under £80 – which is astonishing really.

It's not our winner mind, because as much as we treasure its value for money, stronger performing cards at higher resolutions are well within modest budgets.

Read: Sapphire HD 4850 review

Sapphire HD 4870 512MB - £100

Sapphire HD 4870 512MB - £100

Reluctantly edging our way into the three-digit price bracket, we're led there by the RV770XT-based Radeon HD 4870.

Only twelve months ago, this was AMD's most expensive single chip card: a flagship model that demanded double the number of gold coins and forced NVIDIA not only to start a fi resale, but revise the GTX260's core design.

When the new DirectX 11 cards arrive in a couple of months to accompany the recently-launched Radeon HD 5870, it will likely tumble again, and that's when things will start to get surreal.

Comparing this to the 1GB version opposite, it's obvious that the mere 512MB of GDDR5 present here holds the card back from fulfilling its full potential.

Read: Sapphire HD 4870 review

Sapphire HD 4870 Toxic - £130

Sapphire HD 4870 Toxic - £130

Since the Toxic edition of the HD 4870 is £30 more expensive than the stock version and it has three key improvements over the latter, it follows that if each one is worth ten pounds of your money, then this is the card to buy.

For a start, it comes with the full fat 1GB of memory, and while it may not allow the card to achieve stratospherically higher scores than the 512MB HD 4870, the gap between this and the HD 4850 is the difference between playable and not playable at high quality settings – even though both are saddled with the same number of ROPs.

It's absolutely worth the investment of ten pounds if you are running a larger monitor. That's helped, of course, by the second advantage of the Toxic.

The elegant heatsink isn't just lovely to look at if you have a window in your case, it cools the card sufficiently that Sapphire has upped the clockspeeds by a notch too – and opens up headroom for you to do the same again.

Read: Sapphire HD 4870 Toxic review

Sapphire HD 4890 - £154

Sapphire HD 4890 - £154

We ought to point out that this feature wasn't sponsored by Sapphire – it's just that as the largest manufacturer and supplier of ATI cards its the one with the largest volumes and as such are the easiest to get hold of.

Therefore, if you're going to pick up an ATI card most of the ones you'll find will be under the Sapphire umbrella.

For these reference design cards, you could insert any vendors' name on the box – the overall performance will be the same. Take this example. It's a plain old red-heatsinked Radeon HD4890.

Based on the RV790XT core, it's little more than an overclocked and slightly tweaked version of the RV770XT that powers the HD 4870.

As expected, then, benchmarks are better than that family of cards, even the Toxic edition on the page before, but not an order of magnitude greater. Certainly never more than 10 per cent, even at the most demanding settings.

Read: Sapphire HD 4890 review

Zotac GTS 250 AMP! - £136

Zotac GTS 250 AMP! - £136

The latest addition to the GeForce family seems to repeat the same model as AMD's HD 4770.

The core features fewer shader processors than the full fat 260s and 275s of this world, but runs at extremely high speeds to compensate.

Actually, though, if you look a little closer, the 250 is yet another rebrand of the 8800/9800, with yet another 150MHz or so piled onto the clockspeed, an extra half gig of memory and a launch price of £130-ish.

The chip must be having a real identity crisis, no longer sure if it's an 8 or 9-series or GT-thing.

Read: Zotac GTS 250 AMP! review

BFG GeForce GTX 260 OC - £130

BFG GeForce GTX 260 OC - £130

You should be aware by now that there are two versions of the GeForce GTX 260 kicking around, with little to differentiate them from each other in terms of nomenclature.

For some reason this one has to be called the 'GTX 260 OC'. The original, which was long phased out but still on sale in some quarters, has just 192 stream shaders in its core, the later revision enabled another quad and took that total to 216.

Otherwise, they're more or less the same GPU. The performance difference is considerable, though, so do avoid the earlier model if you can.

Needless to say, this is one of the more powerful cards. Build-wise, the 260 is the same size as the 275, which is just another way of saying 'gigantic'.

Read: BFG GeForce GTX 260 OC review

The winner: XFX GeForce GTX 275XXX - £158

XFX GeForce GTX 275XXX - £158

AIl this talk of price performance ratios is getting a little wearing, but we've saved the best until last.

You cannot get better benchmarks than this for under £160, and in some circumstances it's faster than the £200+ GTX 280.

The only advantage the latter card has is its four extra ROPs, which help with high image quality settings.

Not content with that, this 'XXX' edition is also factory overclocked, albeit by a fairly minor 37MHz. It does mean that realistically, if you want to go beyond this level you're into the realms of CrossFire and SLI multi-chip set-ups and all the driver issues that brings with it.

Like the GTX 260 it's big, but quiet, and based on the 55nm revision of the G200 core that all the top end GeForce cards share.

In the most demanding tests we ran it through, it was over 10 per cent quicker than its nearest competitor, the HD 4890, yet cost just £4 more.

Read: XFX GeForce GTX 275XXX review

How to dual-boot or virtualise Windows 7

Try Windows 7 without affecting your current system

Virtual PC can create virtual machines to host many operating systems

Windows 7 is almost here, and the beta has had such glowing reviews that you might be tempted to try it for yourself.

You could just install it over your existing Windows installation and hope that everything runs smoothly, but that's not recommended, especially because clean installs will be required for European Windows upgraders.

A safer option is to install Windows 7 on a separate hard drive partition so that you can choose to boot either to your existing operating system or to Windows 7 when your PC starts.

Alternatively, using virtualisation software, you could run Windows 7 inside a window on your current XP or Vista desktop, which would be a very convenient way to try it out.

Both approaches have their own quirks and problems, though, so it pays to understand them properly before trying to install Windows 7 on your system. Here's everything that you need to know.

Prepare your machine

Both the dual-boot and virtual machine approaches to running Windows 7 will require a large chunk of your hard drive space: around 10GB when finished and even more for the installation process. These requirements will only grow as you install and use your applications.

If this is an issue for you, take an hour or two to free up hard drive space on your system, running the Windows Disk Cleanup tool, uninstalling unwanted applications and deleting large files that you just don't need any more. It's really not worth trying to cram Windows 7 into a tiny space.

Once that's finished, defragment your hard drive. This consolidates your free space into one or two large blocks, which will help to improve performance, particularly if you're running Windows 7 in a virtual machine. A virtual hard drive that's badly fragmented can really slow you down.

Another complication with the virtual machine approach is the way it chews up system memory. If you want to run Windows 7 in a window on your Vista desktop, you'll need to dedicate at least 1GB of RAM to it if you want to see reasonable performance. Ideally you'd want even more if you're expecting a slick response.

So, even if you have 4GB available you could find around half your memory being grabbed by the Windows 7 virtual machine, which could have a serious effect on your current applications. Now would be a very good time to ensure that your PC is making the best possible use of its RAM.

Browse through all those icons in your system tray, and if you find any apps you don't need, look for options to prevent them loading at boot time. Next, use a tool such as Autoruns to inspect any other programs that are configured to start when Windows loads, and remove those services that are surplus to requirements.

Running Windows 7 in a virtual machine

Microsoft Virtual PC is a free tool used to create and manage virtual machines (read our short review of it). These are essentially software implementations of PC hardware that run in a window on your current desktop. The virtual machine (or VM) is completely isolated from the host PC, so you can install a new operating system and apps, and even format the virtual hard drive without messing up your existing system. Sounds great, right?

VIRTUAL PC: MS Virtual PC has a decent list of supported guest operating systems

Well, it's certainly handy, but there's one significant problem, especially with regard to Windows 7: graphics. VMs are rarely up to high-performance graphics tasks, and so emulate extremely basic video cards that can't handle the full Aero interface. Windows 7 will work, but you'll only see its 'Basic' interface: no transparency, no Aero peek and no big taskbar preview windows.

It's a similar story elsewhere, too: you can run 2D apps like Microsoft Office just fine, but games or other 3D apps almost certainly won't work. If you can live with that limitation, grab a copy of Virtual PC from the Microsoft website. Don't worry too much about the overly fussy system requirements. Microsoft doesn't support running Virtual PC on Windows XP Home, for instance, but in our experience it works just fine.

Windows 7 isn't yet listed as a supported guest operating system either, but we've yet to see any problems. So go ahead, download and install the program – it's surprisingly small and will only take a moment.

Create your virtual system

It's time to roll up our sleeves and get stuck in. Launch Virtual PC and click 'Next' twice to begin the process of creating a virtual machine. Give it a meaningful name like 'Windows 7' and click 'Next'. Choose 'Windows Vista' as your target operating system (it doesn't matter that this is actually incorrect) and click 'Next'.

By default, Virtual PC will allocate 512MB of RAM to this VM, but this amount really isn't enough. We would suggest allocating half your PC's installed RAM, up to a maximum of 1.5GB. To do this, click 'Adjusting the RAM' and choose the figure you'd like before finally clicking 'Next' again.

MEMORY: Make sure you give your Windows 7 enough RAM to operate

Your VM's hard drive will actually just be a large file on your PC's hard drive. Choose to create 'a new virtual hard drive' and click 'Next', then enter the drive size. Enter 25,000MB for now: Virtual PC won't claim this all at once, and you can edit the figure later. Click 'Next', then select 'Finish'.

Now place the Windows 7 disc into a DVD drive, double-click your new VM in the Virtual PC console and wait for the set-up program to load. If it doesn't start up properly, open the CD menu and tell Virtual PC to use the drive containing your disc, then click 'Action | Reset'. The machine should now reboot and launch Windows setup.

Now work your way through the set-up process, choosing the 'Custom (advanced)' install type and pointing Windows 7 at the virtual drive you created earlier. It'll then go to work installing Windows 7 for you, and you can begin trying it out.

How to dual-boot Windows 7

Virtual machines are convenient, but as we've seen, they don't support 3D games or other graphics-intensive applications. There are significant limitations with hardware, too: a Virtual PC VM won't recognise USB devices, for instance (apart from a mouse or keyboard).

And of course you can't properly assess Windows 7's boot time, file copy speed or any other performance measurement by running it in a VM, because it's an artificial environment that will always run significantly slower than a full PC. If you want the full Windows 7 experience, your best option is to create a separate hard drive partition where the operating system can be installed.

When your PC boots you'll then see a new menu with the option to launch either Windows 7 or your current OS. Choose the one you need and it'll have complete access to your hardware with none of the limitations of a VM. If you're running Windows Vista then you can partition your hard drive directly from the Disk Management console. This shouldn't cause any problems at all, but there are no guarantees, so back up your hard drive before you start.

Then launch 'diskmgmt.msc', right-click your current hard drive partition, select 'Shrink volume' and choose the size you'd like to allocate to Windows 7. Again, this should be 25GB at an absolute minimum, and ideally a lot more. Give Windows 7 as much drive space as you can spare.

Now right-click the unallocated space, select 'New simple volume', and work through the wizard, allocating all of the available space to your new partition and performing a quick format with the NTFS filesystem. And that's it: the partitioning is done. Partitions on XP If you have Windows XP, life is a little more complicated, as you'll need to download third-party software to carve up your drive.

Fortunately, there are plenty of free and very capable options. GParted is one of the best out there. Download the GParted ISO disc image and use it to create a bootable CD. Your disc-burning software should have an option to handle ISO images, but if you have problems then use a third party tool like ISO Recorder, which is available from here.

GPARTED: GParted provides an easy way to create and resize XP hard drive partitions

Before you continue, make sure you've backed up your hard drive, or at least any critical data on it. GParted is one of the most heavily used and respected drivepartitioning tools around, so it's most unlikely that your hardware will come to harm, but any lowlevel drive tinkering of this sort has some level of risk attached. As always it's better to be safe than sorry with this sort of thing.

When you're ready to go, boot from the GParted CD, choose the 'Default settings' option and then select 'Don't touch keymap'. Next, pick your language and accept the default operating mode, then locate the partition you're going to resize, select 'Resize/Move' and choose how much space you'd like to give Windows 7 (remember, we suggest 25GB or more).

After that, click 'Resize', then 'Apply'. Next, right-click over the unallocated space and select 'New' from the dropdown menu to see the Create New Partition dialog. Make sure both 'Free space' values are set to zero and that the 'New size' value is set to the maximum partition size, so your new partition is using all the available space.

PARTITION SIZE: Give your partition enough space to fit Windows 7 on easily or you could run into troubles

Choose 'Primary partition' from the 'Create As' list, select 'NTFS' as your filesystem and click 'Add' then 'Apply' to create a Windows 7-sized space on your hard drive. Job done.

Install Windows 7

Now boot your PC from a Windows 7 DVD and work through the installation process. Be sure to choose the 'Custom (advanced)' installation type, then point the set-up program to your empty new partition.

Because you're installing Windows 7 to your real hard drive as opposed to a virtual machine, be very careful here. Click the wrong partition and your current Windows files will be replaced by Windows 7 – not what you wanted at all.

Once you've passed this stage, there's very little more that you need to do. The Windows 7 setup program requires fewer clicks than previous versions, and it gets its work done more quickly too, so a dual–boot installation could be finished in 15 to 30 minutes. You can then start Windows 7 from your PC boot menu, or the Virtual PC console, and start discovering exactly what all the fuss is about.

Virtual machine vs dual-boot

So which Windows 7 trial method is best, then? In most cases, we prefer virtual machines. They're slower and don't fully support 3D graphics or USB devices, but in general that really doesn't matter. We think most people want to explore Windows 7 to try out the new taskbar, libraries, Start menu, search tools or Explorer, not play Crysis or watch Freeview on a USB TV tuner.

If that sounds like you, then it makes sense to try Windows 7 in a virtual machine before you do anything else. However, if after trying the virtual machine you find that you want a more complete look at Windows 7 – or if you know that you'll be frustrated by the lack of functionality within a VM – give dual-boot a try.

It's surprisingly easy, particularly if you're running Vista, and you could have a system running in under 45 minutes.